Biological Problem

Dose-dependent control of gene expression

Many cellular processes are not simple ON/OFF switches. They depend on how much of a given protein or RNA is present:

  • dosage compensation of X-linked genes,
  • haploinsufficiency,
  • fine-tuned thresholds in pluripotency,
  • graded responses to morphogen gradients.

In these settings, changing a transcription factor by 20–30% can be enough to:

  • keep cells pluripotent,
  • push them into a specific lineage,
  • or trigger compensation mechanisms on entire chromosomes.

To understand such systems, you need tools that:

  1. Tune endogenous gene expression within a physiological range,
  2. Do so homogeneously across cells (analog, not digital),
  3. Are reversible and quantitatively characterisable.

CasTuner: analog tuning instead of digital switching

CasTuner is built around degron-controlled Cas-based repressors:

  • A FKBP12^F36V degron domain that makes the repressor level tunable by dTAG-13.
  • Repressors such as hHDAC4–dCas9 (transcriptional) or CasRx (post-transcriptional).
  • Endogenous reporters tagged with fluorescent proteins (e.g. Esrrb-P2A-mCherry, STAG2–EGFP, Nanog–P2A-mCherry).

By titrating dTAG-13, you titrate repressor abundance, which in turn titrates target gene expression. Importantly:

  • KRAB-based systems tend to behave digitally (bimodal: ON vs OFF).
  • hHDAC4–dCas9 and CasRx behave analogically: intermediate repressor levels lead to stable intermediate expression levels at single-cell resolution.

This makes CasTuner an ideal system to:

  • map dose–response curves between transcription factors (e.g. NANOG, OCT4) and their targets,
  • measure kinetic responses to degrader addition/withdrawal,
  • quantify single-cell noise in analog repression,
  • and design new constructs with desired dynamic range, speed and noise.

Role of this Python port

The original CasTuner analysis was implemented in R. The Python port:

  • re-implements the full modelling and analysis logic,
  • uses Snakemake for workflow management,
  • exposes all steps as transparent, inspectable scripts,

  • and produces a final summary report linking:

    • experimental time courses,
    • fitted parameters (half-times, Hill K and n, delays, degradation rates),
    • ODE simulations,
    • noise statistics,
    • and a model-driven design space map.

The rest of this documentation walks through how each step of the pipeline answers a concrete biological question.